Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[Timorous]

Why would someone need an NDA for an internal document.

Where I work the business updates are under NDA and I don't get access to the presentation, the slide deck is shared via video / screen share rather than giving people the document itself.

 

[Markfw]

I mean I just posted this, but no one is using Bergamo for just raw MT perf. Oh sure, technically, Bergamo even beats Genoa in raw MT perf, but again, even AMD is calling these cloud processors, not HPC processors. The reason why is pretty obvious.

<snip>

In which case we get Turin being 48% more performant than GNR, which doesn't shrink the gap between how much SPR loses to Genoa... but it drastically improves the efficiency gap since both of these CPUs will be at 500 watts. Me personally? I think the gap is going to be closer to ~30%, but we will see.
(final disclaimer, ik this is a massively oversimplified projection based on leaks, not including stuff like memory bandwidth or SMT vs 1T, etc etc, but this was just for fun anyway

I think this is basically what I am saying. AMD rules servers in every respect today, and for the foreseeable future. I saw the post about the IT manager thinking "I will never get fired for buying Intel", but I am so sick of that. I can not believe for over 5 years, and with power getting more and more expensive that they won't rethink. They could get a raise if they told their manager "look, this is 30% more efficient on power and is faster". Not to mention in data center, the amount is arguable, but 1 watt saved = 2-3 watts saved due to AC and to APS support.

 

[coercitiv]

Likelihood of increasing IPC while decreasing power is low.

Zen 3 vs. Zen 2

• 19% IPC increase
• same node class, but improved
• slightly higher clocks
• bigger die
• ISO power
• right in our face

Spoiler

The biggest problem in this thread isn't this discussion point though, but rather whether folks around here are going to accept the rude verdicts of a poster as gospel or demand the minimum of proof and decorum.

 

[Thunder 57]

Ok, look at this this way: does it makes sense that all those monumental changes to the core would result in 10-15% IPC jump versus Zen 4? Especially given how much IPC AMD has managed to get with much less transistor investment in previous iterations. It does not make any sense, right?

I'm sure I missed it, but have you given your offical Zen 5 prediction yet?

 

[DisEnchantment]

Massive increase of IPC (Information per Clock) from RGT is the main surprise here.
Normally RGT IPC is pretty poor, 10 seconds of useful info in 10 mins video.

Unlike CPUs however, RGT IPC is susceptible to much more branch mispredicts, "not final BIOS", "last minute design changes", "I have been told", "I could be wrong", etc

 

[gdansk]

I do not want either of those. I can open a directory and sort by last modified date more quickly than I can type the 'advanced copilot' question. And timeline information has been sent to Microsoft since Windows 10, so who knows if they'd even use local processing for that 'feature'.

If these are the killer features perhaps it is best AMD didn't waste die space on NPUs outside of business machines.

 

[StefanR5R]

Intel has a better IPC,

Better IPC in what vs what chip ?

More importantly, better IPC in which workload? If this is not stated or can't be inferred from context, the term "IPC" is useless bullshit.

AMD vs Intel: Which CPUs Are Better in 2023?

We put Intel vs AMD in a battle of processor prowess.

www.tomshardware.com

Here you go.

There is no valid discussion of "IPC" in this article.

But at least it's a fun read. "For professionals on the hunt for performance in content creation and productivity applications, the winner of AMD vs Intel CPUs goes to Intel on the strength of its x86 hybrid architecture." :-D As if.

 

[Schmide]

So You are being intelectually dishonest here, going from "worked fine" to 'certainly knew what they were getting into".

The irony

 

[rtxtwt]

The leak was Kepler agreeing that mass production has started but didn’t say when. What if it was already in production last year?

Possibly. But it usually leaks out when that starts, and this time we’ve not heard anyhing until now.

I posted this 3 months ago

Page 160 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 160 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

few months ago Zen5 DT completion had been already planned to be Oct-Nov, mass production could happen at this timeframe,

but I'm not sure about the release date. My guess is CPU usually being released earlier than APU, GraniteRidge might be earlier than Strixpoint as well. Source is the Greymon(s) and his factory friends and they're my only source, I didn't get any info from others when comes to AMD product. They're still my best reliable source.

 

[Hans Gruber]

Would someone qualify for me a Conroe moment?
Is it like a Zen moment? Or merely A14 sized?

The Conroe moment was when Intel destroyed all of AMD's hopes and dreams for more than a decade. The massive Conroe OC's out of the box that went on over the Intel product cycles for years. I still keep a drawer with my favorite Intel and AMD CPU's.

 

[Gideon]

I suppose this is confirmation of Z5 being the highest IPC core in the world.
Pretty clear that they have made a major breakthrough on the front end, conveniently that is the one area of the core all leaks have said nothing about.
Only shame is about the SMT yield. The fact the thing clocks comfortably over 5GHz is what will make the ARM boys head scratch.

Just to be sure, There aren't any public leaks about this yet, right? As far as the plebs are concerned, this is still only forum / twitter rumor-mill?

Why I ask, is that being a chronical Eastern European I love being needlessly sceptical and wrong, instead of being too hopeful and end up disappointed.

Case in point:

I was super hyped about RDNA3, once this Skyjuice leak surfaced, as it seemed quite obviously real and it was! These decent architectural gains on paper, excellent process shrink (TSMC N5) and promises by the vendor (50% perf/watt improvement) seemed to indicate a slam-dunk. Especially as AMD delivered on RDNA1 and RDNA2 perf/watt promises and had excellent track-record with chiplets.

Turns out, even awesome looking specs & architecture details don't mean much until they are backed up by actual benhmark numbers.

April can't come soon enough.

 

[DisEnchantment]

And so will come the day of the first Zen 5 wielder, when half of the leakers will vanish in oblivion. Perfect balance, restored.

Until a new Zen 6 thread comes, then leakers come out of grinding in the woods fully levelled up bashing level 0 NPCs in new thread.

 

[Abwx]

You should pay more attention: I keep on claiming aggregated scores are an evil thing. And Cinebench is completely useless to me as I don't do rendering. So does that mean Cinebench is a bad benchmark? No. Just that it doesn't mean anything to my (and many people) needs.


Since I'm the one who made the claim, here is multicore speedup for 7950x.

https://browser.geekbench.com/v6/cpu/4915769 Ray Tracer scaling 18.6
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/cinebench-2024-released.2614738/post-41068411 CB24 Scaling 16.7

Doesn't look that bad to me.

Not bad but GB6 is assuming that softwares where it scale badly wont prpgress at all in a 5 years outlook, because i dont think that people update their PC every 2-3 years, so if one look at GB6 to choose a CPU you can be almost sure that it will be outdated in 5 years at most and even earlier if there s some improvement in softwares.

Also even with badly scaling loads their point is that users are performing only this single task at a given time, they dont even account for background tasks or eventually for multiple instances of a badly scaling soft.

 

[AMDK11]

That's pretty raw cope.
Solutions and very large milestones either translate into large IPC increases or you're ARM Austin/Sofia. And you suck.

that's not how semicon pathfinding even works.
the new bits of Zen5 have like, 0 relation to previous Zens, that's the point.
They're not learnings, but novel crackpot concepts of doom.

When writing about research and teaching throughout the entire period of work on Zen 1-4, I mean that not all developed solutions were implemented, but left to subsequent generations.

A microarchitecture that will result in a huge average IPC increase is not necessarily a breakthrough or milestone for AMD. AMD has no choice but to expand the core to enable further IPC gains. At the same time, they implement new techniques.
Therefore, they covered the wider core with new techniques, which in itself is a milestone for AMD.

To achieve an average IPC gain of 30%, the core would need to be expanded (transistors) by approximately 50-60%. Nothing comes for free, and I don't think new techniques will suddenly result in huge increases in IPC without a large increase in transistor counts. ALU itself does not cost that many transistors. Simply adding 50% more ALUs does not necessarily result in 25-30% higher IPC in most workloads in the first generation of a wider core (chip area trade-offs + limited design time).

I understand you have high expectations for the Zen5's mid-range IPC gain. I think you will be disappointed. If it's +25-30% higher IPC on average, I'll be very impressed.

At the moment, I don't have such high expectations, because there has practically never been such a large average jump in x86 IPC from generation to generation. If anything, it's only because formally the previous generation had weak IPC.

Until strong evidence and verification emerges, I maintain my view that Zen 5 will offer an IPC curve ranging from +1-3% to +40-50% with an average I see as a target of +15-20%.

 

[igor_kavinski]

Straight from the horse's mouth (you may call him AMD):

5950X: TSMC 7nm FinFET
7950X: TSMC 5nm FinFET

5900HS: TSMC 7nm FinFET
6900HS: TSMC 6nm FinFET
7940HS: TSMC 4nm FinFET
8945HS: TSMC 4nm FinFET

5900HX: TSMC 7nm FinFET
6900HX: TSMC 6nm FinFET
7945HX: TSMC 5nm FinFET


5850U: TSMC 7nm FinFET
6850U: TSMC 6nm FinFET
7840U: TSMC 4nm FinFET
8840U: TSMC 4nm FinFET

7545U (with Zen4c cores): TSMC 4nm FinFET

 

[Fjodor2001]

Based on past track record, I think adroc is a false prophet. No comment from him regarding why his original April 2024 release date for Zen5 DT is no longer valid. No comment regarding whether Computex 2024 is the new release date for in store availability, or just an announcement.

Also, no comment regarding if the 30+ % IPC increase claim is still valid, and if so using what performance benchmark.

I’m expecting him to continue making claims in this thread, and then disappear once AMD finally releases Zen5 DT officially.

Maybe he’ll then resurface in the Zen6 thread (likely using different user name), and the process will start all over. I’m expecting a new fan crowd, and lots of dead certain claims without any solid evidence.

 

[yuri69]

...when Zen 5 finally launches...

What do you mean "finally"? As AMD stated, Zen release cadence was 12-18 months and Zen 4 had been delayed due COVID and CXL 1.1. This means Zen 5 launches in April. AMIRITE?

 

[eek2121]

…oh I want to add, this thread IS about speculation. Everyone is entitled to voice their own thoughts/opinions as long as they are within the forum guidelines.

 

[DrMrLordX]

Well said... I for one am more than happy to spread the holy words of St Adroc far and wide for fake internet points and updoots.

Let's teach him the fusion dance so he can merge with Nosta and Richie Rich and create the ultimate insane hype machine.

Zen5 is a CON derivative with hybrid CMT/SMT4!!!! You heard it here first people. And you heard it here last, too.

 

[adroc_thurston]

So what happens to Zen 4?

It exists.

Do they stop making Zen 4 chips immediately or are they supposed to slot in below the Zen 5 SKUs for more pricing options for consumers?

DT/mobile Zen4 dies yeah.
Server, Genoa will live for some time on.

 

[Kepler_L2]

Ok, and what was the claimed performance increase for that benchmark, and which CPUs where being compared?

Core for core Zen5 is >40% faster than Zen4 in SPEC.

 

[AMDK11]

IPC increase for Zen 5 > 40% What does this refer to? Is this the average of the IPC growth curve? Up to 40-50%?

It's easy to throw around such numbers, but if there is no reference point and what type of load these numbers refer to, they don't tell us anything. It is equally easy to write that Zen 5 has an IPC >6%. And what does this tell us? Nothing.

Zen 3 with one specific workload (SPEC?) saw +106% compared to Zen 2. However, in the case of AMD's presentation, the IPC growth curve ends at +39% with an average of +19%.

Zen 4, also according to AMD's presentation, the IPC growth curve ends at +39%, but the average is already +13%.

 

[leoneazzurro]

According to this Anandtech review:

AMD Zen 4 Ryzen 9 7950X and Ryzen 5 7600X Review: Retaking The High-End

www.anandtech.com

The average SPECINT 2017 ST increase from Zen3 to Zen4 is 23% considering the average of all substests, considering the platform differences there is probably some lower margin due to the cores only because of memory BW advantage of Zen4, which should not exist with Zen5.

 

[eek2121]

Every time I've tried to access this forum, it has been down FYI, so I'm like way behind in reading. I'm gonna comment on a few things, then I gotta get to bed. Long day today. Long day tomorrow. I also really want to work on this raspberry pi project, it is KILLING me.

Yeaaaa let's just say IDC needs something far meaner than LNC to compete.

They have something, but it isn't ready yet.

BTW, what happened @adroc_thurston and @Kepler_L2 that caused you to change your predictions ~10% higher than just a few weeks ago?:

ARL-S and Zen 5 Desktop Speculation

RULE RULE This spreadsheet is unmoderated. No spamming and please respect what others have written. I will stop making these public speculation spreadsheets once someone breaks the rule. That's all, have fun. (By the way, the tabs are at the bottom.)

docs.google.com

That google sheet was put up on March 9:

https://twitter.com/x/status/1766467520729375119

Zen 5 Hype Train taking off from the station?

He has explained his predictions already. Scroll up.

I didn't see the explanation, but among other things posting exact numbers is never a good idea. There are ways of tracing certain information, but more important, numbers are meaningless without context. That is why I told people to take the Arrow Lake slide for what it is worth.

Our resident Prince Harry expects Arrow Lake to hit 7 GHz!

Intel will RULE!!!

Rule the power infrastructure of a modern country, perhaps.

if Zen 5 is indeed >40% uplift in ST SPECint (which is usually very close to Geekbench ST and the average IPC AMD reports) then this is quite a unique achievement.

The industry standard is about 20% IPC uplift for a new generation (usually on a new process with plenty more transistors).

I am sorry, this is not specifically about Zen 5, but AMD literally did this just a few years ago. Apple did it with the M1 vs both x86 and ARM. Why is it so hard to believe? Because Intel isn't doing it? Intel has been stuck and spinning their wheels to get anywhere. Don't get me wrong, they ARE working on some stuff, but they were far too reliant on node improvements and that absolutely kicked them in the rear end.

Qualcomm, AMD, and Apple have dropped or are dropping chips between this year and next year that are going to bring large improvements to the table. It doesn't happen every year, yes, but it happens frequently. Graviton, for example.

I am also skeptical of the high claims applying to desktop where clock rates are already absurd. But there is no business sense in nerfing a halo product. Even if it is hard to acquire the brand reputation is worth it to have a desktop part pushed to the limits.

It's the reason why some gamers with a measly 4070 will call Radeon people 'the poors' despite those with a 7900 XTX having much better performance. Like the RTX 4090 the 8950X(3D) doesn't have to be available in large numbers or be affordable it simply has to exist to get prestige sales further down the stack.

See above for the first part.

I am a gamer. I own thousands of PC games, hundreds of console games, etc. I don't refer to people like that. I actually would have bought the 7900XTX if it had better RT performance and AMD reference cards were smaller/better because it hit just the right amount of performance/price point for me. Don't get me wrong, when I buy a GPU I like to go all in, but AMD's flagships are always so compelling to me. The 6900XT was pretty amazing, for example. Ultimately I went with a 4090 because it was smaller, quieter, and runs faster. Half the time I game I don't even hear the fans spin up. It also has amazing RT performance. If AMD fixes RT and can focus on delivering a quality reference card (because face it: OEMs don't remember how to innovate) I will get one.

Anyway, I will finish catching up on other threads tomorrow. You guys have a great night. Some of these comments had me giggling.

 

[H433x0n]

Unhappy with Cinebench speculation?
Then let me show you some happy data?

9950X

GeekBench 6
single 3628 more than 23868

GeekBench 5
single 2715 more than 27712

(Only test data, does not represent the final result)

If that’s a legit score that’d be a bad result since a 7950X will regularly score 3000-3200 and a 13900K will usually get a score in the range of 3100-3300.

A 30% 1T increase would mean a score >=4000.